summaryrefslogtreecommitdiffstats
path: root/encoder.c
diff options
context:
space:
mode:
authorDaniel Baumann <daniel.baumann@progress-linux.org>2022-02-21 12:25:28 +0000
committerDaniel Baumann <daniel.baumann@progress-linux.org>2022-02-21 12:25:28 +0000
commite5389ef44a4a53f0694c75c4e3d2b9e64ac3ab2c (patch)
treed35b6bb5cadbc2a260a65bc4ad2806b60d93489d /encoder.c
parentAdding upstream version 1.12. (diff)
downloadclzip-e5389ef44a4a53f0694c75c4e3d2b9e64ac3ab2c.tar.xz
clzip-e5389ef44a4a53f0694c75c4e3d2b9e64ac3ab2c.zip
Adding upstream version 1.13.upstream/1.13
Signed-off-by: Daniel Baumann <daniel.baumann@progress-linux.org>
Diffstat (limited to '')
-rw-r--r--encoder.c131
1 files changed, 59 insertions, 72 deletions
diff --git a/encoder.c b/encoder.c
index ceb8a97..cf68614 100644
--- a/encoder.c
+++ b/encoder.c
@@ -1,5 +1,5 @@
/* Clzip - LZMA lossless data compressor
- Copyright (C) 2010-2021 Antonio Diaz Diaz.
+ Copyright (C) 2010-2022 Antonio Diaz Diaz.
This program is free software: you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
@@ -33,32 +33,25 @@ CRC32 crc32;
int LZe_get_match_pairs( struct LZ_encoder * const e, struct Pair * pairs )
{
- int32_t * ptr0 = e->eb.mb.pos_array + ( e->eb.mb.cyclic_pos << 1 );
- int32_t * ptr1 = ptr0 + 1;
- int32_t * newptr;
- int len = 0, len0 = 0, len1 = 0;
- int maxlen = 3; /* only used if pairs != 0 */
- int num_pairs = 0;
- const int pos1 = e->eb.mb.pos + 1;
- const int min_pos = ( e->eb.mb.pos > e->eb.mb.dictionary_size ) ?
- e->eb.mb.pos - e->eb.mb.dictionary_size : 0;
- const uint8_t * const data = Mb_ptr_to_current_pos( &e->eb.mb );
- int count, key2, key3, key4, newpos1;
- unsigned tmp;
int len_limit = e->match_len_limit;
-
if( len_limit > Mb_available_bytes( &e->eb.mb ) )
{
len_limit = Mb_available_bytes( &e->eb.mb );
if( len_limit < 4 ) return 0;
}
- tmp = crc32[data[0]] ^ data[1];
- key2 = tmp & ( num_prev_positions2 - 1 );
+ int maxlen = 3; /* only used if pairs != 0 */
+ int num_pairs = 0;
+ const int min_pos = ( e->eb.mb.pos > e->eb.mb.dictionary_size ) ?
+ e->eb.mb.pos - e->eb.mb.dictionary_size : 0;
+ const uint8_t * const data = Mb_ptr_to_current_pos( &e->eb.mb );
+
+ unsigned tmp = crc32[data[0]] ^ data[1];
+ const int key2 = tmp & ( num_prev_positions2 - 1 );
tmp ^= (unsigned)data[2] << 8;
- key3 = num_prev_positions2 + ( tmp & ( num_prev_positions3 - 1 ) );
- key4 = num_prev_positions2 + num_prev_positions3 +
- ( ( tmp ^ ( crc32[data[3]] << 5 ) ) & e->eb.mb.key4_mask );
+ const int key3 = num_prev_positions2 + ( tmp & ( num_prev_positions3 - 1 ) );
+ const int key4 = num_prev_positions2 + num_prev_positions3 +
+ ( ( tmp ^ ( crc32[data[3]] << 5 ) ) & e->eb.mb.key4_mask );
if( pairs )
{
@@ -67,7 +60,7 @@ int LZe_get_match_pairs( struct LZ_encoder * const e, struct Pair * pairs )
if( np2 > min_pos && e->eb.mb.buffer[np2-1] == data[0] )
{
pairs[0].dis = e->eb.mb.pos - np2;
- pairs[0].len = maxlen = 2;
+ pairs[0].len = maxlen = 2 + ( np2 == np3 );
num_pairs = 1;
}
if( np2 != np3 && np3 > min_pos && e->eb.mb.buffer[np3-1] == data[0] )
@@ -86,18 +79,23 @@ int LZe_get_match_pairs( struct LZ_encoder * const e, struct Pair * pairs )
}
}
+ const int pos1 = e->eb.mb.pos + 1;
e->eb.mb.prev_positions[key2] = pos1;
e->eb.mb.prev_positions[key3] = pos1;
- newpos1 = e->eb.mb.prev_positions[key4];
+ int newpos1 = e->eb.mb.prev_positions[key4];
e->eb.mb.prev_positions[key4] = pos1;
+ int32_t * ptr0 = e->eb.mb.pos_array + ( e->eb.mb.cyclic_pos << 1 );
+ int32_t * ptr1 = ptr0 + 1;
+ int len = 0, len0 = 0, len1 = 0;
+
+ int count;
for( count = e->cycles; ; )
{
- int delta;
if( newpos1 <= min_pos || --count < 0 ) { *ptr0 = *ptr1 = 0; break; }
- delta = pos1 - newpos1;
- newptr = e->eb.mb.pos_array +
+ const int delta = pos1 - newpos1;
+ int32_t * const newptr = e->eb.mb.pos_array +
( ( e->eb.mb.cyclic_pos - delta +
( (e->eb.mb.cyclic_pos >= delta) ? 0 : e->eb.mb.dictionary_size + 1 ) ) << 1 );
if( data[len-delta] == data[len] )
@@ -152,7 +150,6 @@ static void LZe_update_distance_prices( struct LZ_encoder * const e )
for( len_state = 0; len_state < len_states; ++len_state )
{
int * const dsp = e->dis_slot_prices[len_state];
- int * const dp = e->dis_prices[len_state];
const Bit_model * const bmds = e->eb.bm_dis_slot[len_state];
int slot = 0;
for( ; slot < end_dis_model; ++slot )
@@ -161,6 +158,7 @@ static void LZe_update_distance_prices( struct LZ_encoder * const e )
dsp[slot] = price_symbol6( bmds, slot ) +
(((( slot >> 1 ) - 1 ) - dis_align_bits ) << price_shift_bits );
+ int * const dp = e->dis_prices[len_state];
for( dis = 0; dis < start_dis_model; ++dis )
dp[dis] = dsp[dis];
for( ; dis < modeled_distances; ++dis )
@@ -169,7 +167,7 @@ static void LZe_update_distance_prices( struct LZ_encoder * const e )
}
-/* Returns the number of bytes advanced (ahead).
+/* Return the number of bytes advanced (ahead).
trials[0]..trials[ahead-1] contain the steps to encode.
( trials[0].dis4 == -1 ) means literal.
A match/rep longer or equal than match_len_limit finishes the sequence.
@@ -178,9 +176,8 @@ static int LZe_sequence_optimizer( struct LZ_encoder * const e,
const int reps[num_rep_distances],
const State state )
{
- int main_len, num_pairs, i, rep, num_trials, len;
- int rep_index = 0, cur = 0;
- int replens[num_rep_distances];
+ int num_pairs, num_trials;
+ int i, rep, len;
if( e->pending_num_pairs > 0 ) /* from previous call */
{
@@ -189,8 +186,10 @@ static int LZe_sequence_optimizer( struct LZ_encoder * const e,
}
else
num_pairs = LZe_read_match_distances( e );
- main_len = ( num_pairs > 0 ) ? e->pairs[num_pairs-1].len : 0;
+ const int main_len = ( num_pairs > 0 ) ? e->pairs[num_pairs-1].len : 0;
+ int replens[num_rep_distances];
+ int rep_index = 0;
for( i = 0; i < num_rep_distances; ++i )
{
replens[i] = Mb_true_match_len( &e->eb.mb, 0, reps[i] + 1 );
@@ -212,10 +211,7 @@ static int LZe_sequence_optimizer( struct LZ_encoder * const e,
return main_len;
}
- {
const int pos_state = Mb_data_position( &e->eb.mb ) & pos_state_mask;
- const int match_price = price1( e->eb.bm_match[state][pos_state] );
- const int rep_match_price = match_price + price1( e->eb.bm_rep[state] );
const uint8_t prev_byte = Mb_peek( &e->eb.mb, 1 );
const uint8_t cur_byte = Mb_peek( &e->eb.mb, 0 );
const uint8_t match_byte = Mb_peek( &e->eb.mb, reps[0] + 1 );
@@ -227,6 +223,9 @@ static int LZe_sequence_optimizer( struct LZ_encoder * const e,
e->trials[1].price += LZeb_price_matched( &e->eb, prev_byte, cur_byte, match_byte );
e->trials[1].dis4 = -1; /* literal */
+ const int match_price = price1( e->eb.bm_match[state][pos_state] );
+ const int rep_match_price = match_price + price1( e->eb.bm_rep[state] );
+
if( match_byte == cur_byte )
Tr_update( &e->trials[1], rep_match_price +
LZeb_price_shortrep( &e->eb, state, pos_state ), 0, 0 );
@@ -250,9 +249,8 @@ static int LZe_sequence_optimizer( struct LZ_encoder * const e,
for( rep = 0; rep < num_rep_distances; ++rep )
{
- int price;
if( replens[rep] < min_match_len ) continue;
- price = rep_match_price + LZeb_price_rep( &e->eb, rep, state, pos_state );
+ const int price = rep_match_price + LZeb_price_rep( &e->eb, rep, state, pos_state );
for( len = min_match_len; len <= replens[rep]; ++len )
Tr_update( &e->trials[len], price +
Lp_price( &e->rep_len_prices, len, pos_state ), rep, 0 );
@@ -272,17 +270,10 @@ static int LZe_sequence_optimizer( struct LZ_encoder * const e,
if( ++len > e->pairs[i].len && ++i >= num_pairs ) break;
}
}
- }
+ int cur = 0;
while( true ) /* price optimization loop */
{
- struct Trial *cur_trial, *next_trial;
- int newlen, pos_state, triable_bytes, len_limit;
- int start_len = min_match_len;
- int next_price, match_price, rep_match_price;
- State cur_state;
- uint8_t prev_byte, cur_byte, match_byte;
-
Mb_move_pos( &e->eb.mb );
if( ++cur >= num_trials ) /* no more initialized trials */
{
@@ -290,8 +281,8 @@ static int LZe_sequence_optimizer( struct LZ_encoder * const e,
return cur;
}
- num_pairs = LZe_read_match_distances( e );
- newlen = ( num_pairs > 0 ) ? e->pairs[num_pairs-1].len : 0;
+ const int num_pairs = LZe_read_match_distances( e );
+ const int newlen = ( num_pairs > 0 ) ? e->pairs[num_pairs-1].len : 0;
if( newlen >= e->match_len_limit )
{
e->pending_num_pairs = num_pairs;
@@ -300,7 +291,8 @@ static int LZe_sequence_optimizer( struct LZ_encoder * const e,
}
/* give final values to current trial */
- cur_trial = &e->trials[cur];
+ struct Trial * cur_trial = &e->trials[cur];
+ State cur_state;
{
const int dis4 = cur_trial->dis4;
int prev_index = cur_trial->prev_index;
@@ -331,25 +323,25 @@ static int LZe_sequence_optimizer( struct LZ_encoder * const e,
mtf_reps( dis4, cur_trial->reps ); /* literal is ignored */
}
- pos_state = Mb_data_position( &e->eb.mb ) & pos_state_mask;
- prev_byte = Mb_peek( &e->eb.mb, 1 );
- cur_byte = Mb_peek( &e->eb.mb, 0 );
- match_byte = Mb_peek( &e->eb.mb, cur_trial->reps[0] + 1 );
+ const int pos_state = Mb_data_position( &e->eb.mb ) & pos_state_mask;
+ const uint8_t prev_byte = Mb_peek( &e->eb.mb, 1 );
+ const uint8_t cur_byte = Mb_peek( &e->eb.mb, 0 );
+ const uint8_t match_byte = Mb_peek( &e->eb.mb, cur_trial->reps[0] + 1 );
- next_price = cur_trial->price +
- price0( e->eb.bm_match[cur_state][pos_state] );
+ int next_price = cur_trial->price +
+ price0( e->eb.bm_match[cur_state][pos_state] );
if( St_is_char( cur_state ) )
next_price += LZeb_price_literal( &e->eb, prev_byte, cur_byte );
else
next_price += LZeb_price_matched( &e->eb, prev_byte, cur_byte, match_byte );
/* try last updates to next trial */
- next_trial = &e->trials[cur+1];
+ struct Trial * next_trial = &e->trials[cur+1];
Tr_update( next_trial, next_price, -1, cur ); /* literal */
- match_price = cur_trial->price + price1( e->eb.bm_match[cur_state][pos_state] );
- rep_match_price = match_price + price1( e->eb.bm_rep[cur_state] );
+ const int match_price = cur_trial->price + price1( e->eb.bm_match[cur_state][pos_state] );
+ const int rep_match_price = match_price + price1( e->eb.bm_rep[cur_state] );
if( match_byte == cur_byte && next_trial->dis4 != 0 &&
next_trial->prev_index2 == single_step_trial )
@@ -364,11 +356,11 @@ static int LZe_sequence_optimizer( struct LZ_encoder * const e,
}
}
- triable_bytes =
+ const int triable_bytes =
min( Mb_available_bytes( &e->eb.mb ), max_num_trials - 1 - cur );
if( triable_bytes < min_match_len ) continue;
- len_limit = min( e->match_len_limit, triable_bytes );
+ const int len_limit = min( e->match_len_limit, triable_bytes );
/* try literal + rep0 */
if( match_byte != cur_byte && next_trial->prev_index != cur )
@@ -392,19 +384,20 @@ static int LZe_sequence_optimizer( struct LZ_encoder * const e,
}
}
+ int start_len = min_match_len;
+
/* try rep distances */
for( rep = 0; rep < num_rep_distances; ++rep )
{
const uint8_t * const data = Mb_ptr_to_current_pos( &e->eb.mb );
const int dis = cur_trial->reps[rep] + 1;
- int price;
if( data[0-dis] != data[0] || data[1-dis] != data[1] ) continue;
for( len = min_match_len; len < len_limit; ++len )
if( data[len-dis] != data[len] ) break;
while( num_trials < cur + len )
e->trials[++num_trials].price = infinite_price;
- price = rep_match_price + LZeb_price_rep( &e->eb, rep, cur_state, pos_state );
+ int price = rep_match_price + LZeb_price_rep( &e->eb, rep, cur_state, pos_state );
for( i = min_match_len; i <= len; ++i )
Tr_update( &e->trials[cur+i], price +
Lp_price( &e->rep_len_prices, i, pos_state ), rep, cur );
@@ -412,17 +405,14 @@ static int LZe_sequence_optimizer( struct LZ_encoder * const e,
if( rep == 0 ) start_len = len + 1; /* discard shorter matches */
/* try rep + literal + rep0 */
- {
int len2 = len + 1;
const int limit = min( e->match_len_limit + len2, triable_bytes );
- int pos_state2;
- State state2;
while( len2 < limit && data[len2-dis] == data[len2] ) ++len2;
len2 -= len + 1;
if( len2 < min_match_len ) continue;
- pos_state2 = ( pos_state + len ) & pos_state_mask;
- state2 = St_set_rep( cur_state );
+ int pos_state2 = ( pos_state + len ) & pos_state_mask;
+ State state2 = St_set_rep( cur_state );
price += Lp_price( &e->rep_len_prices, len, pos_state ) +
price0( e->eb.bm_match[state2][pos_state2] ) +
LZeb_price_matched( &e->eb, data[len-1], data[len], data[len-dis] );
@@ -435,21 +425,19 @@ static int LZe_sequence_optimizer( struct LZ_encoder * const e,
e->trials[++num_trials].price = infinite_price;
Tr_update3( &e->trials[cur+len+1+len2], price, rep, cur + len + 1, cur );
}
- }
/* try matches */
if( newlen >= start_len && newlen <= len_limit )
{
- int dis;
const int normal_match_price = match_price +
price0( e->eb.bm_rep[cur_state] );
while( num_trials < cur + newlen )
e->trials[++num_trials].price = infinite_price;
- i = 0;
+ int i = 0;
while( e->pairs[i].len < start_len ) ++i;
- dis = e->pairs[i].dis;
+ int dis = e->pairs[i].dis;
for( len = start_len; ; ++len )
{
int price = normal_match_price + LZe_price_pair( e, dis, len, pos_state );
@@ -502,7 +490,7 @@ bool LZe_encode_member( struct LZ_encoder * const e,
int price_counter = 0; /* counters may decrement below 0 */
int dis_price_counter = 0;
int align_price_counter = 0;
- int ahead, i;
+ int i;
int reps[num_rep_distances];
State state = 0;
for( i = 0; i < num_rep_distances; ++i ) reps[i] = 0;
@@ -539,7 +527,7 @@ bool LZe_encode_member( struct LZ_encoder * const e,
Lp_update_prices( &e->rep_len_prices );
}
- ahead = LZe_sequence_optimizer( e, reps, state );
+ int ahead = LZe_sequence_optimizer( e, reps, state );
price_counter -= ahead;
for( i = 0; ahead > 0; )
@@ -556,14 +544,13 @@ bool LZe_encode_member( struct LZ_encoder * const e,
const uint8_t prev_byte = Mb_peek( &e->eb.mb, ahead + 1 );
const uint8_t cur_byte = Mb_peek( &e->eb.mb, ahead );
CRC32_update_byte( &e->eb.crc, cur_byte );
- if( St_is_char( state ) )
+ if( ( state = St_set_char( state ) ) < 4 )
LZeb_encode_literal( &e->eb, prev_byte, cur_byte );
else
{
const uint8_t match_byte = Mb_peek( &e->eb.mb, ahead + reps[0] + 1 );
LZeb_encode_matched( &e->eb, prev_byte, cur_byte, match_byte );
}
- state = St_set_char( state );
}
else /* match or repeated match */
{